RESUMO
BACKGROUND: A large number of web-based COVID-19 symptom checkers and chatbots have been developed; however, anecdotal evidence suggests that their conclusions are highly variable. To our knowledge, no study has evaluated the accuracy of COVID-19 symptom checkers in a statistically rigorous manner. OBJECTIVE: The aim of this study is to evaluate and compare the diagnostic accuracies of web-based COVID-19 symptom checkers. METHODS: We identified 10 web-based COVID-19 symptom checkers, all of which were included in the study. We evaluated the COVID-19 symptom checkers by assessing 50 COVID-19 case reports alongside 410 non-COVID-19 control cases. A bootstrapping method was used to counter the unbalanced sample sizes and obtain confidence intervals (CIs). Results are reported as sensitivity, specificity, F1 score, and Matthews correlation coefficient (MCC). RESULTS: The classification task between COVID-19-positive and COVID-19-negative for "high risk" cases among the 460 test cases yielded (sorted by F1 score): Symptoma (F1=0.92, MCC=0.85), Infermedica (F1=0.80, MCC=0.61), US Centers for Disease Control and Prevention (CDC) (F1=0.71, MCC=0.30), Babylon (F1=0.70, MCC=0.29), Cleveland Clinic (F1=0.40, MCC=0.07), Providence (F1=0.40, MCC=0.05), Apple (F1=0.29, MCC=-0.10), Docyet (F1=0.27, MCC=0.29), Ada (F1=0.24, MCC=0.27) and Your.MD (F1=0.24, MCC=0.27). For "high risk" and "medium risk" combined the performance was: Symptoma (F1=0.91, MCC=0.83) Infermedica (F1=0.80, MCC=0.61), Cleveland Clinic (F1=0.76, MCC=0.47), Providence (F1=0.75, MCC=0.45), Your.MD (F1=0.72, MCC=0.33), CDC (F1=0.71, MCC=0.30), Babylon (F1=0.70, MCC=0.29), Apple (F1=0.70, MCC=0.25), Ada (F1=0.42, MCC=0.03), and Docyet (F1=0.27, MCC=0.29). CONCLUSIONS: We found that the number of correctly assessed COVID-19 and control cases varies considerably between symptom checkers, with different symptom checkers showing different strengths with respect to sensitivity and specificity. A good balance between sensitivity and specificity was only achieved by two symptom checkers.
Assuntos
Infecções por Coronavirus/diagnóstico , Infecções por Coronavirus/epidemiologia , Autoavaliação Diagnóstica , Internet , Pneumonia Viral/diagnóstico , Pneumonia Viral/epidemiologia , Avaliação de Sintomas/instrumentação , Adolescente , Adulto , Algoritmos , Betacoronavirus , COVID-19 , Teste para COVID-19 , Centers for Disease Control and Prevention, U.S. , Técnicas de Laboratório Clínico , Coleta de Dados , Humanos , Pessoa de Meia-Idade , Pandemias , Valor Preditivo dos Testes , Informática em Saúde Pública , Reprodutibilidade dos Testes , SARS-CoV-2 , Autorrelato , Sensibilidade e Especificidade , Estados Unidos , Adulto JovemRESUMO
Objective: We retrospectively screened 350,116 electronic health records (EHRs) to identify suspected patients for Pompe disease. Using these suspected patients, we then describe their phenotypical characteristics and estimate the prevalence in the respective population covered by the EHRs. Methods: We applied Symptoma's Artificial Intelligence-based approach for identifying rare disease patients to retrospective anonymized EHRs provided by the "University Hospital Salzburg" clinic group. Within 1 month, the AI screened 350,116 EHRs reaching back 15 years from five hospitals, and 104 patients were flagged as probable for Pompe disease. Flagged patients were manually reviewed and assessed by generalist and specialist physicians for their likelihood for Pompe disease, from which the performance of the algorithms was evaluated. Results: Of the 104 patients flagged by the algorithms, generalist physicians found five "diagnosed," 10 "suspected," and seven patients with "reduced suspicion." After feedback from Pompe disease specialist physicians, 19 patients remained clinically plausible for Pompe disease, resulting in a specificity of 18.27% for the AI. Estimating from the remaining plausible patients, the prevalence of Pompe disease for the greater Salzburg region [incl. Bavaria (Germany), Styria (Austria), and Upper Austria (Austria)] was one in every 18,427 people. Phenotypes for patient cohorts with an approximated onset of symptoms above or below 1 year of age were established, which correspond to infantile-onset Pompe disease (IOPD) and late-onset Pompe disease (LOPD), respectively. Conclusion: Our study shows the feasibility of Symptoma's AI-based approach for identifying rare disease patients using retrospective EHRs. Via the algorithm's screening of an entire EHR population, a physician had only to manually review 5.47 patients on average to find one suspected candidate. This efficiency is crucial as Pompe disease, while rare, is a progressively debilitating but treatable neuromuscular disease. As such, we demonstrated both the efficiency of the approach and the potential of a scalable solution to the systematic identification of rare disease patients. Thus, similar implementation of this methodology should be encouraged to improve care for all rare disease patients.